Previous news story    Next news story

Adobe admits using 'synthetic blur' image in deblur demo

By dpreview staff on Oct 18, 2011 at 22:00 GMT

Adobe has admitted an image used in its 'image deblur' presentation was artificially blurred for the purposes of the demonstration. The company said the blur on the image was 'more complicated than anything we can simulate using Photoshop's blur capabilities.' It described the move as 'common practice in research' and defended the use of the image because 'we wanted it to be entertaining and relevant to the audience.' The other images shown were the result of camera shake, it said.

The original, unshaken, image of Adobe's Kevin Lynch by Kendall Whitehouse

The admission came as part of an article on the company's Photoshop Blog, that describes the technology in more detail.

Adobe's statement on the issue, from the Photoshop Blog:

'The first two images we showed – the crowd scene and the image of the poster, were examples of motion blur from camera shake. The image of Kevin Lynch was synthetically blurred from a sharp image taken from the web. What do we mean by synthetic blur? A synthetic blur was created by extracting the camera shake information from another real blurry image and applying it to the Kevin Lynch image to create a realistic simulation.'

'This kind of blur is created with our research tool. Because the camera shake data is real, it is much more complicated than anything we can simulate using Photoshop’s blur capabilities. When this new image was loaded as a JPEG into the deblur plug-in, the software has no idea it was synthetically generated. This is common practice in research and we used the Kevin example because we wanted it to be entertaining and relevant to the audience – Kevin being the star of the Adobe MAX conference!'

Comments

Total comments: 97
12
Mssimo
By Mssimo (Oct 18, 2011)

Does not sound like they were cheating. I wonder if this deblur function would correct bokeh (or just make it very ugly)

2 upvotes
gurgeh
By gurgeh (Oct 18, 2011)

It won't be able to correct out of focus areas - the blurring has to be directional. It will still be interesting to see how it will affect OOF areas, because that was not clear from the examples they shown.

0 upvotes
tkbslc
By tkbslc (Oct 18, 2011)

So it would have been too hard to take a shot of Kevin at 1s handheld to use for the example?

7 upvotes
cordellwillis
By cordellwillis (Oct 18, 2011)

Yeah I was wondering the same. We're talking digital not film, so why on earth couldn't they just use an image from a digicam, cell, tablet, other?

1 upvote
Joseph S Wisniewski
By Joseph S Wisniewski (Oct 19, 2011)

Any random handheld shot is unlikely to be useful for this sort of demo. The motion in the example shots is a long path, with a dwell (hence the ghost image in the shopping mall shot). A more realistic motion blur would oscillate, crossing back over itself several times, and the deconvolution won't work anywhere near as well on such an image.

5 upvotes
Ulfric M Douglas
By Ulfric M Douglas (Oct 19, 2011)

I strongly suspect they 'cheated' on the other images too, although in a less obvious way.
If you can create a 'camera shaken' picture with very very simple camera movement you can 'unblur' it convincingly. Anything else ... maybe not. I'd like to see detailed proof of origin of the two 'genuine' photos.

0 upvotes
GBo
By GBo (Oct 19, 2011)

Hi all, I tried to unblur the "crowd" image and it worked :
http://gbotet.perso.sfr.fr/deblur/guess.jpg
In PNG full size :
http://gbotet.perso.sfr.fr/deblur/exampleOut.png

I used the program found here :
http://www.cse.cuhk.edu.hk/~leojia/programs/deblurring/deblurring.htm
More info here (incl.a paper):
http://www.cse.cuhk.edu.hk/~leojia/projects/motion_deblurring/index.html

=> there are still some artefacts, but that's maybe because I did not try to refine parameters except the kernel size : 17x17

This deblur.exe program is using the same kernel for all points of the image, that's means that the motion blur is uniform, i think that it is called "spatially invariant motion blur" in this research field, right ?
Is it a realistic/average case of camera shake blur ?
I don't know about Adobe's plug-in capabilities, but for sure the crowd image they used can be perfectly deblurred with only one kernel. Simple case I guess.

regards,
GBo

Comment edited 2 times, last edit 1 minute after posting
2 upvotes
Total comments: 97
12